If you are looking for LinkedIn lookalike audiences, they no longer exist. LinkedIn discontinued them on February 29, 2024: new ones cannot be created, and old ones are frozen as static audiences that stop refreshing. The replacement is predictive audiences, which you build from your own source data and then size with a slider. That slider is where most accounts go wrong. A predictive audience only copies the seed you feed it, and the wider you set it, the less it resembles that seed.
What replaced lookalikes
LinkedIn points to two tools. Predictive audiences are created on the Audiences page in Campaign Manager from one source: a contact list, a company list, a conversion, a Lead Gen Form, or a retargeting audience. The source needs at least 300 members, and LinkedIn's AI builds an audience predicted to act like it. Audience Expansion is a checkbox at ad set level that widens reach to members similar to your targeting, and LinkedIn positions it for awareness rather than conversion. The two do not stack: Audience Expansion is turned off on any ad set that uses a predictive audience.
Size is a function of the seed, not the slider
A predictive audience is capped by its source. Seed it from a large, loose list and even a tight setting inherits loose fit. Seed it from your actual converters (a conversion or Lead Gen Form source, or a CRM list of closed-won accounts uploaded as a matched audience) or a clean ICP list, and a modest audience stays on target. Fix the seed before you touch the size.
Wider reach is not free
Every notch wider on the slider adds people who match the seed less well and cost the same to reach. On LinkedIn, where clicks are already premium, that dilution shows up fast as rising cost per qualified lead. LinkedIn suggests campaign audiences of at least 50,000 members, so start near that floor, prove the audience converts, and widen only while lead quality holds. Give it time before you judge: LinkedIn says a new predictive audience can take up to five days to reach its full size, and it refreshes daily after that.
Where it fits
Treat the predictive audience as one rung: a clean converter seed feeds it, firmographic targeting sanity-checks it, and ABM lists cover the named accounts it will never perfectly capture. Reach is not the goal; reaching more of the right people is. Meta still has lookalikes, and sizing a Meta lookalike follows the same seed-first logic.
Common questions
Does LinkedIn still have lookalike audiences?
No. LinkedIn discontinued lookalike audiences on February 29, 2024. New ones cannot be created, and existing ones are frozen as static audiences that no longer refresh. LinkedIn's replacements are predictive audiences and Audience Expansion.
What is the best size for a LinkedIn predictive audience?
There is no single best size; the seed sets it. A predictive audience built from real converters or a clean ICP list stays on target at a modest size, while a loose seed produces a loose audience at any size. LinkedIn suggests campaign audiences of at least 50,000 members, so start near that floor and move the slider wider only while lead quality holds.
Does a bigger predictive audience perform better?
Not usually. Moving the size slider wider adds people who resemble the seed less, at the same cost per click, which tends to raise cost per qualified lead. Reach is not the goal; fit is.
What is the difference between predictive audiences and Audience Expansion?
Predictive audiences are built on the Audiences page from your own source data (a contact list, company list, conversion, Lead Gen Form, or retargeting audience) and aim at people likely to convert. Audience Expansion is a checkbox at ad set level that widens reach to similar members and is aimed at awareness. LinkedIn turns Audience Expansion off on ad sets that use a predictive audience.